template-benchmark

JMH benchmark for popular Java template engines:

Running the benchmark

Download the source code and build it (mvn clean install)
(Optional) To run a benchmark for a single template, such as Mustache, use java -jar target/benchmarks.jar Mustache
Run the benchmark for presentations or stocks workload. E.g. for : presentations

java -jar target/benchmarks.jar -i 4 -wi 4 -f 1 -r 2 -w 2 -rff results.csv -rf csv -tu ms presentations

Generating plot

Run Stocks benchmark while exporting results to csv with java -jar target/benchmarks.jar -rff results.csv -rf csv stocks
Use gnuplot to generate plot with gnuplot benchmark.plot. This will output results.png.
Run Presentations benchmark while exporting results to csv with java -jar target/benchmarks.jar -rff results.csv -rf csv presentations
Use gnuplot to generate plot with gnuplot benchmark.plot. This will output results.png.

Presentations benchmark was imported from spring-comparing-template-engines.

Rules of Template Engine Configuration

It is imperative that each template engine is configured in way to reflect real-world usage as opposed to it's optimal configuration. Typically this means an out-of-the-box configuration.

To strive for a common set of features across template engines, the following configurations are expected:

Disabling of HTML escaping
Template loaded from classpath prior to actual benchmark

Interpreting the Results

The benchmarks measure throughput, given in "ops/time". The time unit used is seconds. Generally, the score represents the number of templates rendered per second; the higher the score, the better.

Example Results (2024, with HtmlFlow 4 and JStachio)

Upgraded all releases of templates including HtmlFlow to 4.3 and add new template engine JStachio

These tests were performed in Github Actions with Ubuntu 22.04, 4 CPU and 16 GB RAM according to GitHub-hosted runners for Public repositories.

Plots are generated from the data collected in this run

Example Results (2023, with HtmlFlow 3.5)

These tests differ from the 2019 results in the upgraded releases of JMH and Kotlin standard libraries from 1.2.61 to 1.8.20 (Kotlin included several optimizations to enhance type safe assertions).

These tests were performed on a local machine with the following specs:

Mac OS Version 13.4.1
Apple MacBook M1 Pro, 8 Cores and 16 GB RAM
OpenJDK Runtime Environment Corretto-17.0.5.8.1 
OpenJDK 64-Bit Server VM Corretto (build 17.0.5+8-LTS)

Example Results (2019, with HtmlFlow 3.5)